hoangsonww/Customizable-AI-Chatbot

Implements a modular intention-detection pipeline that classifies user queries before routing to appropriate response handlers, with RAG-powered question answering that uses Hypothetical Document Embeddings (HyDE) to retrieve contextual chunks from Pinecone before generating responses. The frontend leverages Shadcn UI and Radix UI primitives with Framer Motion for animations, while the Next.js backend supports streaming responses via multiple LLM providers with configurable fallback logic and includes GitHub Actions CI/CD automation.